Cyp26 enzymes function in endoderm to regulate pancreatic field size.

Abstract

The control of organ size and position relies, at least in part, upon appropriate regulation of the signals that specify organ progenitor fields. Pancreatic cell fates are specified by retinoic acid (RA), and proper size and localization of the pancreatic field are dependent on tight control of RA signaling. Here we show that the RA-degrading Cyp26 enzymes play a critical role in defining the normal anterior limit of the pancreatic field. Disruption of Cyp26 function causes a dramatic expansion of pancreatic cell types toward the anterior of the embryo. The cyp26a1 gene is expressed in the anterior trunk endoderm at developmental stages when RA is signaling to specify pancreas, and analysis of cyp26a1/giraffe (gir) mutant zebrafish embryos confirms that cyp26a1 plays the primary role in setting the anterior limit of the pancreas. Analysis of the gir mutants further reveals that cyp26b1 and cyp26c1 function redundantly to partially compensate for loss of Cyp26a1 function. We used cell transplantation to determine that Cyp26a1 functions directly in endoderm to modulate RA signaling and limit the pancreatic field. Taken together with our finding that endodermal expression of cyp26 genes is subject to positive regulation by RA, our data reveal a feedback loop within the endoderm. Such feedback can maintain consistent levels of RA signaling, despite environmental fluctuations in RA concentration, thus ensuring a consistent size and location of the pancreatic field.

Full text

Page 1

Cyp26 enzymes function in endoderm to regulate
pancreatic field size
Mary D. Kinkela,1, Elizabeth M. Seftona, Yutaka Kikuchib, Takamasa Mizoguchib, Andrea B. Wardc,
and Victoria E. Princea,d,1
aDepartment of Organismal Biology and Anatomy anddCommittee on Developmental Biology, University of Chicago, 1027 East 57th Street, Chicago, IL
60637;bDepartment of Biological Science, Graduate School of Science, Hiroshima University, Kagamiyama 1-3-1, Higashi-Hiroshima, Hiroshima 739-8526
Japan; andcBiology Department, Adelphi University, 1 South Avenue, Garden City, NY 11530
Edited by Donald F. Steiner, University of Chicago, Chicago, IL, and approved March 19, 2009 (received for review December 22, 2008)
The control of organ size and position relies, at least in part, upon
appropriate regulation of the signals that specify organ progenitor
fields. Pancreatic cell fates are specified by retinoic acid (RA), and
proper size and localization of the pancreatic field are dependent
on tight control of RA signaling. Here we show that the RA-
degradingCyp26enzymesplayacriticalroleindefiningthenormal
anterior limit of the pancreatic field. Disruption of Cyp26 function
causes a dramatic expansion of pancreatic cell types toward the
anterior of the embryo. The cyp26a1 gene is expressed in the
anterior trunk endoderm at developmental stages when RA is
signaling to specify pancreas, and analysis of cyp26a1/giraffe (gir)
mutant zebrafish embryos confirms that cyp26a1 plays the primary
role in setting the anterior limit of the pancreas. Analysis of the gir
mutants further reveals that cyp26b1 and cyp26c1 function redun-
dantly to partially compensate for loss of Cyp26a1 function. We
used cell transplantation to determine that Cyp26a1 functions
directly in endoderm to modulate RA signaling and limit the
pancreatic field. Taken together with our finding that endodermal
expression of cyp26 genes is subject to positive regulation by RA,
our data reveal a feedback loop within the endoderm. Such
feedback can maintain consistent levels of RA signaling, despite
environmental fluctuations in RA concentration, thus ensuring a
consistent size and location of the pancreatic field.
?-cell ? insulin ? retinoic acid ? zebrafish
D
axis to form distinct tissue types or organs in the appropriate
location. Endoderm regionalization is influenced by a variety
of signals from the adjacent mesoderm, including retinoic acid
(RA) (reviewed in ref. 1). When zebrafish RA signaling is
disrupted pharmacologically or genetically, no pancreas de-
velops (2). There is a similar requirement for RA to specify
pancreas in avians and amphibians (3, 4), and development of
the mouse dorsal pancreatic bud is also dependent on RA
signals (5, 6). Conversely, treatment of zebrafish gastrulae with
exogenous RA expands the number and domain of pancreas
cells into the anterior endoderm (2). Timed treatments with
the RA receptor inhibitor BMS493 have established that RA
specifies the zebrafish pancreas between 9 and 13 h postfer-
tilization (hpf), at the end of gastrulation. At these stages the
aldh1a2/neckless (nls) gene, encoding the RA synthetic en-
zyme Raldh2, has high expression levels in the anterior
paraxial mesoderm. Cell transplantation studies have con-
firmed that this anterior mesoderm is the essential source of
RA signals to specify pancreas, and that these signals are in
turn received directly in the adjacent prepancreatic endoderm
(7). Because RA plays a critical role in promoting pancreas
specification, the development of a pancreatic field with
appropriate size and location is in turn dependent on proper
regulation of RA signaling during normal development.
We ask here whether the Cyp26 retinoic acid 4-hydroxylase
enzymes play a role in regulating RA signaling in the endoderm
uring vertebrate development the endoderm germ layer
becomes regionalized along its anterior–posterior (AP)
to control pancreas size and position. The cyp26 genes encode
cytochrome P450 enzymes that degrade retinoids to their inac-
tive hydroxylated polar derivatives (reviewed in ref. 8). Targeted
disruptions of mouse Cyp26a1 cause a suite of defects that
resemble those induced by excess RA (9, 10), and Niederreither
et al. (11) found that this phenotype could be partially rescued
by lowering RA signaling via heterozygous mutation of Aldh1a2.
In zebrafish, as in mice, there are 3 members of the cyp26 gene
family—cyp26a1, cyp26b1, and cyp26c1—that function to limit
the levels of RA signaling in the embryo (12–15). The 3 zebrafish
cyp26 genes play redundant functions in the developing hind-
brain, where they regulate RA-dependent gene expression (15).
Reduction of zebrafish Cyp26 activity, using the cyp26a1 mutant
giraffe (girrw716), by morpholino knockdown or using pharmaco-
logical inhibitors, leads to excess RA signaling and hence
changes in hindbrain patterning (13, 15, 16). Similarly, double
mutant analysis of mouse Cyp26a1 and Cyp26c1 has indicated
redundant roles in hindbrain patterning. Cyp26 function has also
been implicated in patterning of the zebrafish pronephros (17)
and of the mouse pharyngeal region (18), but function of cyp26
genes in the developing postpharyngeal digestive tract has not
previously been investigated.
Here we demonstrate that cyp26a1 is expressed in the ze-
brafish anterior trunk endoderm at the appropriate develop-
mental stage and location to regulate RA-dependent pancreas
specification. Consistent with this model, we find that blocking
Cyp26 function significantly enlarges the pancreatic field toward
the anterior. Cyp26a1 plays the primary role in this process, but
in the absence of functional Cyp26a1, the cyp26b1 and cyp26c1
genescanpartiallycompensateforitsloss.WhileCyp26enzymes
could potentially act as a ‘‘sink’’ for diffusible RA by action in
any germ layer, we find it is endodermal Cyp26a1 function that
regulates the pancreatic field. Finally, endodermal Cyp26 ex-
pression is itself positively regulated by RA signals. This implies
a feedback loop to ensure that, despite likely fluctuations in RA
levels, there is uniform RA signaling and hence a consistently-
sized pancreatic field.
Results
RA Signaling Positively Regulates Expression of cyp26 Genes in the
Anterior Trunk Endoderm. To identify endodermal genes that are
positively regulated by RA signaling, we performed microarray
transcription profiling comparing gene expression levels be-
tween wild-type endoderm and RA-deficient or RA-treated
Author contributions: M.D.K., E.M.S., A.B.W., and V.E.P. designed research; M.D.K., E.M.S.,
and A.B.W. performed research; Y.K. and T.M. contributed new reagents/analytic tools;
M.D.K., E.M.S., A.B.W., and V.E.P. analyzed data; and M.D.K. and V.E.P. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1To whom correspondence may be addressed. E-mail: mkinkel@uchicago.edu or
vprince@uchicago.edu.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0813108106/DCSupplemental.
7864–7869 ?
PNAS ?
May 12, 2009 ?
vol. 106 ?
no. 19www.pnas.org?cgi?doi?10.1073?pnas.0813108106

Page 2

endoderm. To facilitate this analysis we purified GFP-labeled
endoderm cells from 10 hpf Tg(sox17:EGFP) embryos (19) using
fluorescence-activated cell sorting (FACS). Hybridization to
Agilent microarrays revealed that cyp26a1 expression was up-
regulated135-foldinresponsetoRAtreatment,andcyp26b1was
up-regulated 14-fold; cyp26c1 is not represented on the array.
This dramatic up-regulation in response to exogenous RA
suggests that endodermal Cyp26 expression is subject to feed-
back regulation, as described for neural ectoderm (15).
We next asked whether endodermal cyp26 is expressed at an
appropriatetimeandlocationtoregulatepancreasdevelopment.
Published studies of cyp26a1 indicate expression at the anterior
and posterior of the embryo, commencing at late blastula stage
(12), with a new domain of expression arising in the anterior
trunk by 13 hpf (17, 20). We analyzed cyp26a1 expression
between 8 and 13 hpf, when RA signaling is acting on pancreatic
progenitors (2). We found that cyp26a1 expression in anterior
trunk endoderm arises between 9 and 10 hpf and gradually
increases in level over time (Fig. 1 A–G). Colabeling with myod,
an adaxial mesoderm and somite marker, indicated that the
cyp26a1 anterior limit of expression corresponds to the level of
the anterior-most paraxial mesoderm (Fig. 1 H and I). Thus
anterior trunk endoderm expression of cyp26a1 correlates ap-
proximately with the location of pancreatic progenitors as
assessed by lineage tracing (21) and by pdx1 pancreas progenitor
marker expression from 15 hpf (1). Additionally, this endodermal
cyp26a1 expression is immediately subjacent to the anterior
somites, which are the site of aldh1a2 expression and are the
source of the RA signals that specify the pancreas (7). In
summary, our expression analysis of cyp26a1 reveals endodermal
expression at approximately the AP location of pancreas pro-
genitors, at the stages when they are being specified.
To confirm our microarray data, we asked whether endodermal
expression of cyp26a1 is up-regulated by RA treatment. As ex-
pected, RA treatment led to a dramatic up-regulation of cyp26a1
expression (Fig. 1 J–N). Sections through the anterior trunk of
mock-treated embryos showed that cyp26a1 expression is primarily
localized to endoderm and ectoderm (Fig. 1M), as in wild types.
Sections through RA-treated embryos confirmed elevated expres-
sion in both ectoderm and endoderm (Fig. 1N). Conversely, in
embryos treated with DEAB to block RA synthesis, the anterior
trunk expression of cyp26a1 was absent (Fig. 1L), demonstrating
that RA signaling is required for establishment of this cyp26a1
expression domain. Our microarray analysis of RA signaling-
deficient embryos did not show a significant reduction in endoder-
mal cyp26a1 expression relative to wild types. This suggests that
endodermalexpressionattherostralandcaudalendsoftheembryo
does not require RA signaling, consistent with the lack of aldh1a2
expression in these regions. We also investigated expression of
cyp26b1 and cyp26c1 between 8 and 16 hpf. We did not detect
cyp26b1 expression in the anterior trunk of wild-type embryos at
these stages (Fig. 1O); rather, expression was confined to the
A
EFGH
J
KL
M
O
P
Q
N
R
I
B
CD
Fig. 1.
views (A–C) and lateral views (D–G). Arrowheads indicate endodermal expression. (H and I) Double in situ hybridizations with cyp26a1 (purple) and myod (red)
to show paraxial mesoderm. Embryos are deyolked and flat-mounted, and anterior is to top of page. Arrowheads indicate anterior limit of myod expression.
(J) cyp26a1 expression in DMSO carrier-treated embryo at 12 hpf, whole-mount, lateral view. (K) cyp26a1 expression in embryo treated with 0.1 ?M RA from
5.25hpf;noteup-regulationofexpression.(L)cyp26a1expressionin10?M4-(diethylamino)benzaldehyde(DEAB)-treated(RA-deficient)embryofrom5.25hpf;
note absence of anterior trunk expression. (M) Transverse optical section through anterior trunk at 12 hpf. (N) Transverse section through anterior trunk of
RA-treated embryo as in K; note up-regulation of endodermal cyp26a1. (O) cyp26b1 expression at 12 hpf, whole-mount, lateral view. (P) cyp26b1 expression in
embryotreatedwith0.1?MRAfrom5.25hpf.(Q)cyp26c1expressionat12hpf,whole-mount,lateralview.(R)Transversesectionthroughanteriortrunk,embryo
treatedwith0.1?MRAfrom5.25hpf;noteup-regulationofendodermalcyp26b1.Arrowheadindicatesanteriortrunk,andforM,N,andRarrowheadsindicates
anterior trunk endoderm.
Expression of cyp26 genes in late zebrafish gastrulae. (A–G) Expression of cyp26a1 from 8 to 13 hpf. Shown are whole mounts, animal pole up, dorsal
Kinkel et al.PNAS ?
May 12, 2009 ?
vol. 106 ?
no. 19 ?
7865
DEVELOPMENTAL
BIOLOGY

Page 3

hindbrain, as reported (15). However, in RA- treated embryos, we
detected robust up-regulation of cyp26b1 in the anterior trunk
endoderm(Fig.1PandR),consistentwithourmicroarrayanalysis.
By contrast, we were unable to detect cyp26c1 expression in the
anterior trunk of wild-type or RA-treated specimens (Fig. 1Q). In
summary,insituanalysisconfirmedthatexpressionofbothcyp26a1
and cyp26b1 is up-regulated in the endoderm in response to
exogenousRAandalsoshowedthatthisup-regulationoccursinthe
domain that correlates with the developing pancreas progenitors.
Inhibition of Cyp26 Function Expands the Pancreatic Field Toward the
Anterior. Because cyp26a1 gene expression overlaps with pan-
creas progenitor cells at the stages when they are being specified
by RA, we asked whether Cyp26 negative regulators of RA
signaling might play a role in limiting the domain over which RA
can induce pancreatic fates. We rendered zebrafish embryos
unable to metabolize RA by treating them with the pan-Cyp26-
inhibitor R115866, and analyzed the expression of endodermal
or pancreatic markers at 24 hpf or later. In contrast to mock-
treated embryos (Fig. 2A), expression of the pancreas progenitor
marker pdx1 was expanded anteriorly within the endoderm of
Cyp26-deficient embryos (Fig. 2B), a result that closely mim-
icked treatment with 0.1 ?M RA (2). We found a corresponding
anterior expansion in the expression of markers of differentiated
pancreas derivatives such as insulin (Fig. 2 C–F), glucagon (Fig. 2
G and H), and somatostatin 2 (Fig. 2 I and J), as well as in the
endodermal domain of the endocrine pancreas progenitor
marker islet1 (Fig. 2 K and L). The liver marker cebpa showed a
similar expansion toward the anterior (Fig. 2 M and N). Similar
to effects of RA treatment, this expansion of foregut derivatives
is at the expense of more anterior endoderm derivatives, as
revealed by the anterior shift in the pharyngeal domain of foxA2
expression (Fig. 2 O and P). As expected, whether R115866 was
included or not, no insulin expression was detected in the
presence of DEAB, consistent with the R115866 phenotype
being a consequence of blocking the ability of Cyp26 enzymes to
degrade RA. Finally, we colabeled 24 hpf R115866-treated
embryos with insulin and an antibody against the muscle marker
Myosin to visualize the somites. This confirmed that the poste-
rior limit of the insulin expression domain was unchanged by
R115866 treatment. In summary, we find that Cyp26 function is
required to suppress RA signaling and thus prevent inappropri-
ate expansion of the pancreatic field toward the anterior.
To ask when Cyp26 is functioning to limit the pancreatic field,
we treated embryos with R115866 for 1-h time windows com-
mencing at hourly intervals between 5 and 15 hpf (Fig. S1).
Treatment at 13 hpf or later did not significantly alter the AP
extent of insulin (ins)-expressing ?-cells, consistent with our
previousfindingthatRAfunctionsbefore13hpf(2).Treatments
commenced between 5 and 11 hpf had similar consequences,
suggestingthatonceappliedR115866maynotberemovable,and
that Cyp26 enzymes normally complete the majority of their
function by 12 hpf. We conclude that Cyp26 enzymes function in
a comparable time frame to RA signaling, although our data do
not allow us to distinguish the precise onset of Cyp26 function.
Mutant and Morpholino Analysis Reveals Redundant Functions of
Cyp26 Enzymes. Because cyp26a1 has a robust expression pattern
in the prepancreatic endoderm, we asked whether R115866
inhibitortreatmentsarephenocopiedbymutationofcyp26a1/gir.
We compared the AP extent of the insulin-expressing domain in
R115866-treated embryos and gir mutants. The gir mutants
showed a significant anterior expansion of the pancreatic do-
main, but both the size of the domain and the number of
insulin-expressing ?-cells were smaller in the mutants than in the
inhibitor-treated embryos (Fig. 3 A, B, and I). This finding
suggests that when Cyp26a1 function is blocked, other cyp26
genes may take over part of its function. To test this hypothesis,
we performed morpholino knockdown of cyp26b1 and/or
cyp26c1 in wild-type and gir embryos. We found that knockdown
of cyp26b1 or cyp26c1, either singly or in combination in the
wild-type background, did not expand the pancreatic domain
(Fig. 3 C–E, I), consistent with the lack of expression of these
genes in the prepancreatic domain of wild-type embryos (Fig. 1).
Similarly, when either was knocked down in the gir mutant, the
pancreas was not expanded farther than expected with loss of
A
C
E
G
B
D
FH
I
K
MO
J
L
NP
Fig. 2.
?-cells. (E and F) insulin (ins), lateral view. (G and H) glucagon (gcga) labels ?-cells. (I and J) somatostatin 2 (sst2) labels ?-cells. (K and L) islet1 labels endocrine
pancreas; lateral view. (M and N) cebpa labels liver and intestine. (O and P) foxa2 labels pharyngeal endoderm. Arrowheads indicate endoderm. The dorsal view
is shown unless otherwise indicated, anterior to left.
Cyp26 inhibition expands the pancreas anteriorly. (A and B) pdx1 labels cells of the pancreas and anterior small intestine. (C and D) insulin (ins) labels
7866 ?
www.pnas.org?cgi?doi?10.1073?pnas.0813108106 Kinkel et al.

Page 4

cyp26a1 alone (Fig. 3 F and G, I). However, knockdown of both
cyp26b1 and cyp26c1 together in the gir mutant recapitulated the
complete R115866 inhibitor treatment (Fig. 3 H and I). This
finding suggests that when Cyp26a1 function is lost, cyp26b1 and
cyp26c1 provide a redundant compensatory mechanism to en-
sure at least a partial limitation of the pancreatic domain. This
model predicts that elevated RA signaling, in response to loss of
Cyp26a1 function, should induce endodermal expression of
cyp26b1 and cyp26c1. Consistent with this prediction, we have
indeed found induction of endodermal cyp26b1 expression in the
anterior trunk in response to RA treatment (Fig. 1). Although
we were unable to detect equivalent induction of endodermal
cyp26c1 expression, our data suggest that functional levels of
Cyp26c1 enzyme are indeed present.
Cell Transplantation Demonstrates That Cyp26a1 Functions Within
Endoderm to Restrict Pancreas Location. Next we asked in which
germ layer Cyp26a1 functions to define the anterior limit of the
pancreatic domain. From early in gastrulation, and through the
stages of pancreas specification, cyp26a1 is expressed in all 3
germ layers (12) (Fig. 1). Our previous studies have demon-
strated that RA synthesized in the paraxial mesoderm signals
directly to receptors in the endoderm to specify pancreas (7). We
hypothesized that Cyp26a1 might function to restrict pancreas
size either by degrading RA at its source, in the paraxial
mesoderm, or by degrading RA at its site of action, in the
endoderm.
To distinguish between these alternative hypotheses, we used
a cell transplantation strategy to knock down Cyp26a1 function
only in specific tissues. To test whether Cyp26a1 functions in the
mesoderm, we manipulated Cyp26a1 expression in anterior
paraxial mesoderm, the source of the RA that specifies pancreas
(7). We transplanted wild-type donor cells to the anterior
paraxial mesoderm of an otherwise Cyp26a1-deficient host
(schematized in Fig. 4A). We found that, as with Cyp26a1
knockdown throughout the embryo, the insulin-positive domain
was expanded in the AP axis at 24 hpf (n ? 5/5, Fig. 4B). In
reciprocal experiments, we transplanted Cyp26a1-deficient do-
nor cells to the anterior paraxial mesoderm of a wild-type host
(schematized in Fig. 4C), and did not find an expanded insulin-
positive domain (n ? 4/5) (Fig. 4D). Although a single embryo
showed an AP expansion of the insulin-positive domain, the cells
were sparse and scattered, suggesting a developmental delay
rather than altered pancreas location (1). Together, these data
suggest that Cyp26a1 activity in paraxial mesoderm does not
influence pancreatic domain size or location. Similarly, we did
not find any alteration to the pancreatic domain of specimens
where transplanted cells contributed to the neural ectoderm.
AlthoughwecannotruleoutarequirementforCyp26a1function
in a broader mesodermal or ectodermal domain, our findings
nevertheless suggest that Cyp26a1 may function within
endoderm to limit RA signaling.
B
A
C
D
E
F
G
H
I
Fig. 3.
Knockdown of Cyp26b1 and Cyp26c1 singly or in combination does not expand insulin. (F and G) Knockdown of Cyp26b1 or Cyp26c1 in gir mutants expands the
pancreas as with gir alone. (H) Deficiency in all 3 Cyp26 enzymes expands the pancreas more than with Cyp26a1 deficiency alone. All panels show 24 hpf, dorsal
view,anteriortotheleft,200?magnification.(I)SizeoftheinsulindomaininresponsetoCyp26deficiency.NotethattheCyp26a1morpholinodoesnotexpand
the pancreas as dramatically as in gir mutants. Wild-type, n ? 19; a1MO, n ? 10; b1MO, n ? 34; c1MO, n ? 28; b1 ? c1 MO, n ? 20; R115866, n ? 10; gir ? b1MO,
n ? 17; gir ? c1MO, n ? 11; gir ? b1 ? c1MO, n ? 11.
Cyp26 enzymes set the anterior limit of the insulin domain. (A) Wild-type insulin domain. (B) insulin is expanded in Cyp26a1-deficient gir mutants. (C–E)
A
B
C
D
E
F
G
H
Fig. 4.
(A,C,E,andG)Celltransplantationstrategy.Atthe1-cellstage,embryoswere
injectedwiththeindicatedreagents.At4hpf(spherestage),donorcellswere
transplanted to host blastoderm margin. (B, D, F, and H) insulin assayed at 24
hpf, dorsal view, anterior to the left, 200? magnification. (B) Wild-type
anteriorsomiticmesodermisnotsufficienttorescuethepancreasphenotype.
(D) Cyp26a1 knockdown in anterior somitic mesoderm does not expand the
pancreas. (F) Cyp26a1 knockdown in endoderm expands the AP insulin do-
main. (H) Wild-type endoderm rescues the pancreas phenotype.
Cyp26a1 functions within the endoderm to limit insulin expression.
Kinkel et al. PNAS ?
May 12, 2009 ?
vol. 106 ?
no. 19 ?
7867
DEVELOPMENTAL
BIOLOGY

Page 5

To test directly whether Cyp26a1 functions in the endoderm,
we transplanted donor cells specifically to the endoderm germ
layer by manipulating Sox32 function (7, 22, 23). Briefly, we
injected donor embryos with Sox32 mRNA, causing all mesen-
doderm progenitors to take on an endoderm fate, and then we
transplanted donor endoderm cells to hosts that were rendered
endoderm-deficient by Sox32 knockdown. Using this approach,
we generated chimeric embryos in which all endoderm was
host-derived (as previously described in ref. 7). Only those
specimens in which donor-derived endoderm contributed to the
full AP extent of the host embryo were used for further analysis.
In control transplants between embryos with intact Cyp26a1
function, we found that the insulin-expressing cells were local-
ized to the usual wild-type AP domain; we noted, however, that
the number of ?-cells was often reduced in endoderm-
transplanted specimens, compared with unmanipulated speci-
mens (1). We then manipulated Cyp26a1 function specifically
within the endoderm by using the transplantation strategy
schematized in Fig. 4E. We found that endoderm-specific knock-
down of Cyp26a1 caused the insulin-positive domain to expand
in the AP axis (n ? 4/4, Fig. 4F), similar to the consequence of
Cyp26a1 knockdown in all 3 germ layers. This result supports the
model that endodermal Cyp26a1 limits RA signaling. To test this
hypothesis further, we asked whether wild-type endoderm cells
could rescue the pancreas phenotype in a Cyp26a1-deficient
host. We found that transplantation of wild-type endoderm to an
otherwise Cyp26a1-deficient host (schematized in Fig. 4G) did
indeed restore the AP extent of insulin expression to the
wild-type domain (n ? 2/2, Fig. 4H). We conclude that Cyp26a1
functionswithintheendodermtorestricttheanteriorlimitofthe
pancreas.
Discussion
The prepancreatic endoderm is patterned by RA that is synthe-
sized in the anterior trunk mesoderm. The RA diffuses to the
underlying endoderm, where it binds to its receptors to regulate
transcription of downstream target genes and thus specify the
pancreas(7).ThemesodermalRAissynthesizedinanAPregion
that is considerably larger than the prepancreatic domain, which
raises the question of how the RA signal is limited to a precise
endodermal domain during pancreas specification. Here we
show that the anterior extent of the pancreas is restricted by
endodermally expressed Cyp26a1, which degrades RA in the
endoderm and thus restricts the anterior extent of the pancreas.
We previously showed that the endodermally expressed Cdx4
transcription factor limits RA signaling in the posterior
endoderm, thus setting the posterior limit of the pancreas (1).
Together, these studies suggest a model in which foregut
endoderm is patterned by the combined action of mesoderm-
derived RA and endodermal factors that restrict its effect. The
size of the pancreatic domain is defined not by limiting RA in
the mesoderm, but rather by restricting RA signaling within the
endoderm. This mechanism ensures that mesodermal RA re-
mains available for signaling to other tissues, such as the
ectoderm, where it has important roles in hindbrain patterning
(13, 15, 16).
We have demonstrated that initiation of expression of cyp26a1
in the anterior trunk domain is dependent upon RA signaling.
This implies that mesoderm-derived RA signals to endoderm
progenitors for a limited time before being negatively regulated
by newly-induced endodermal Cyp26a1. Such time-dependent
RA signaling has been suggested to underlie aspects of zebrafish
hindbrain patterning (8). Because exogenous RA expands the
cyp26a1 expression domain within the trunk, but not throughout
the embryo, it is likely that this gene is subject to complex
regulation; this can be addressed by functional studies of cyp26a1
regulatory elements.
The RA-degrading activity of the Cyp26 enzymes is an
important mechanism by which the embryo can correct for
fluctuating levels of RA in the environment (reviewed in ref. 8).
Stores of RA are deposited maternally into the yolk, as are stores
of dietary vitamin A, the precursor of RA (24, 25). These stores
vary between embryos because the amount of vitamin A present
inthematernaldietfluctuates,andtherateofdepositionintothe
yolk varies, as has been shown for numerous bird species (e.g.,
ref. 26 and references therein); thus a means to refine RA
signaling levels is critical to ensure normal development. Here
we have shown that normal pancreas development relies on
Cyp26a1 function to control the size of the pancreas and to set
its anterior limit. The existence of a redundant back-up mech-
anism, provided by Cyp26b1 and Cy26c1, further demonstrates
the importance of setting limits to RA activity within endoderm.
This study may be of relevance to the development of proto-
cols for differentiating ?-cells from stem cells. A standard in the
field is to expose undifferentiated cells to a progressive admin-
istration of signaling molecules, including RA, to step cells
through a differentiation program that mimics the in vivo
differentiation of prepancreatic endoderm to functional ?-cells
(27–29). The rationale of this approach is that protocols to
differentiate stem cells should take cues from the normal
embryonic signaling environment. Here we have demonstrated
that during normal embryonic development endoderm cells
express Cyp26 inhibitors of RA signaling to limit ?-cell differ-
entiation. This finding suggests that the efficiency of ?-cell
differentiation from stem cells may be improved by reducing
function of Cyp26 enzymes.
Materials and Methods
Endoderm Sorting and Microarray Analysis. Tg[?5.0sox17:EGFP]zf99 embryos
were injected with raldh2 morpholino (GeneTools) or treated with 1 ?M RA
(Sigma) between 8 and 9 hpf (as described in ref. 7). Embryos were grown to
10 hpf, and 30–50 embryos were dissociated in Hanks solution containing
0.15% trypsin (Sigma) and 2.4 units/mL dispase (Gibco). Cells were pelleted at
400 ? g for 5 min and resuspended in 1 mL of Hanks solution with 5% BSA.
GFP-positive endoderm cells were collected by FACS, and RNA was isolated
andamplifiedasdescribed(30)thenhybridizedtotheAgilentzebrafisholigo
microarray V1. The entire experiment was performed in triplicate.
Pharmacological Treatments and Analysis. Wild-type embryos were treated
with 10 ?M DEAB (Aldrich), from 5.25 hpf until fixation; or with 10 ?M
R115866 (Janssen Pharmaceutica), from 4.33 hpf until fixation (unless other-
wise noted); or with 0.1 ?M RA from 5.25 hpf until fixation. All reagents were
diluted in DMSO and treatments were performed in the dark.
FortimedR115866treatments,embryoswereincubatedinthedarkfor1h,
then washed 3 times with embryo medium and fixed at 24 hpf. Controls were
treated with 0.1% DMSO from 5 hpf until fixation. To measure the insulin
domain, embryos were deyolked, flat-mounted, and imaged using a Zeiss
Axioscope. AP length of the expression domain was measured using Adobe
Photoshop and calibrated using a micrometer.
Microinjection, Cell Transplantation, and in Situ Hybridization. Morpholinos for
cyp26a1 (13), cyp26b1 (15), and cyp26c1 (15) were used at 5 ng/?L. Cell
transplantation was performed as described (7); in situ hybridization was
performed as described (31), using the following probes: aldh1a2 (32), cebpa
(33), cyp26a1 (12), cyp26b1 (14), cyp26c1 (34), foxa2 (35), glucagon (36), hhex
(37), insulin (38), islet1 (39), myod1 (40), pdx1 (41), and somatostatin 2 (36).
Myosin antibody staining was performed as previously described (1).
ACKNOWLEDGMENTS. We thank Cecilia Moens (Fred Hutchinson Cancer
ResearchCenter,Seattle),TetsuKudoh(UniversityofExeter,Exeter,UK),Tom
Schilling (University of California, Irvine, CA), Betsy Dobbs-McAuliff (Central
Connecticut State University, New Britain, CT), and Elwood Linney (Duke
University, Durham, NC) for kindly providing Cyp26 probes and morpholinos;
Jessen Pharmaceutica for providing R115866; Xinmin Li, of The University of
Chicago Microarray Facility, for assistance with microarray performance and
analysis; Ryan Duggan of The University of Chicago FACS facility for assistance
with endoderm cell sorting; and Isaac Skromne and Oni Mapp for comments on
the manuscript. The myosin monoclonal antibody developed by Helen M. Blau
was obtained from the Developmental Studies Hybridoma Bank and was devel-
oped under the auspices of the National Institute of Child Health and Human
7868 ?
www.pnas.org?cgi?doi?10.1073?pnas.0813108106Kinkel et al.